Three decades of deforestation in southwest Sumatra:
Have protected areas halted forest loss and logging,
and promoted re-growth?
David L.A. Gaveaua,b,*, Hagnyo Wandonoc, Firman Setiabudid
a
Wildlife Conservation Society-Indonesia Program, Jl. Pangrango No. 8, Bogor, Indonesia
Durrell Institute of Conservation and Ecology, Department of Anthropology, University of Kent, Canterbury, CT2 7NR, Kent, UK
c
Bukit Barisan Selatan National Park Office, Directorate General of Forest Protection and Nature Conservation (PHKA),
Kota Agung Barat, Lampung Province, Indonesia
d
Illegal Logging Response Centre-European Union/Ministry of Forestry, Manggala Wanabakti Bld. Block VII, 6th Floor,
Jl. Jend. Gatot Subroto, Jakarta, Indonesia
b
A R T I C L E I N F O
A B S T R A C T
Article history:
Much of the forest cover in southern Sumatra, Indonesia has been cleared since the early
Received 7 February 2006
1970s, but accurate estimates of the scales and rates of loss are lacking. This study com-
Received in revised form
bined high-quality remote sensing applications and extensive field surveys, both to provide
5 August 2006
an accurate picture of deforestation patterns across an area of 1.17 million ha in southwest
Accepted 31 August 2006
Sumatra and to assess whether southwest Sumatra’s Bukit Barisan Selatan National Park
Available online 1 November 2006
(BBSNP) has halted forest loss and logging, and promoted re-growth, since its creation in
1984. Of the single large (692,850 ha) contiguous area of forest standing across our study
Keywords:
area in 1972, nearly half (344,409 ha) has been cleared from 1972 to 2002, at an average rate
Protected Areas (PA)
per original forest cover of 1.69% y
Forest loss
Hydrological Reserves (HR), forests have shrunk by 28,696 ha and 113,105 ha, at an average
Logging
rate of 2.74% y
Forest re-growth
four times more slowly than those in GRWS and HR, and have shrunk by 57,344 ha, at an
Satellite imagery
average rate of 0.64% y
Southwest Sumatra
rapidly during the post-establishment, as during the pre-establishment, period (0.65% y
and 0.63% y
1
and 2.13% y
1
. In Gunung Raya Wildlife Sanctuary (GRWS) and
1
, respectively. In contrast, forests in BBSNP have reduced
1
. Nevertheless, the forests within BBSNP were cleared almost as
1
1
, respectively) despite the introduction of protection measures during the
post-establishment period, following the government’s pledge to expand and protect Indonesia’s network of Protected Areas (PAs) at the 1982 Bali World Parks Congress. While these
protection measures failed to slow down rates of forest loss caused by agricultural
encroachments they reduced large-scale mechanised logging by a factor of 4.2 and stabilized some 8610 ha of agricultural encroachments, enabling forest re-growth.
Ó 2006 Elsevier Ltd. All rights reserved.
* Corresponding author: Address: Wildlife Conservation Society-Indonesia Program, Jl. Pangrango No. 8, Bogor, Indonesia. Tel.: +62
251342135.
E-mail addresses: [email protected], [email protected] (D.L.A. Gaveau), [email protected] (H. Wandono), ilrc-gis@
cyber-isp.net.id (F. Setiabudi).
0006-3207/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.biocon.2006.08.035
1.
Introduction
Despite the dramatic loss of the world’s humid tropical forests
(Achard et al., 2002), an interview-based study in 93 tropical terrestrial Protected Areas (PAs) world-wide suggests that >83% of
the sites surveyed are maintaining or increasing their forest
cover (Bruner et al., 2001). Several satellite-based studies concur with the conclusions of Bruner et al. (2001), and indicate
that tropical terrestrial PAs are generally effective at slowing
deforestation within their boundaries (Sanchez-Azofeifa
et al., 2003; Defries et al., 2005; Naughton-Treves et al., 2005;
Nepstad et al., 2006). Nevertheless, DeFries et al. (2005) noted
that Southeast Asian PAs inhibit forest loss less effectively than
Latin America PAs. In Indonesian Borneo (Kalimantan) protected lowland forests declined by more than 56% during the
period 1985 to 2001, mainly due to concession-based timber
extraction and plantation establishment (Curran et al., 2004).
In Indonesia’s second largest island (Sumatra), where forests
declined by 6.7 million ha from 1985 to 1997 (FWI/GFW, 2002)
the effectiveness of PAs at conserving forest habitat is also
being questioned (Jepson et al., 2001; Kinnaird et al., 2003; Linkie et al., 2004). However, the lack of accurate estimates of the
scales and rates of forest loss within Sumatran PAs has limited
research on PA effectiveness in this part of the tropical world.
Here, we present a novel empirical approach to test the effectiveness of Sumatran PAs at conserving forest habitat. Our
approach employs high-quality satellite-based data and extensive field surveys, rather than interviews (Bruner et al., 2001)
that are vulnerable to the biases of informants (Nepstad et al.,
2006). The method incorporates deforestation rates in and
around PAs (Defries et al., 2005; Naughton-Treves et al., 2005;
Nepstad et al., 2006) and deforestation rates before and after
PA establishment (Liu et al., 2002). It also includes re-growth
and logging mitigation that are rarely discussed in the literature
on PA effectiveness. We examined data for the Bukit Barisan
Selatan National Park (BBSNP) in Southern Sumatra because
of its important areas of lowland tropical evergreen forest. In
1972, the BBSNP and two other reserves of lesser protection status protected parts of a single, large contiguous area of forest.
First, we determined the fragmentation and deforestation rates
of this forested area. Second, we compared the fragmentation
and the rates of forest loss across unprotected forests and protected forests, to test whether unprotected forests have reduced
more rapidly than protected forests and whether forests within
PAs of lesser protection status have reduced more rapidly than
forests within the BBSNP. Third, we measured the rates of forest
loss before and after the BBSNP was declared a national park, to
test whether the current BBSNP prevented agricultural
encroachments. Fourth, we measured logging trail networks
during BBSNP’s pre- and post-establishment phases, to test
whether the current BBSNP prevented large scale mechanised
logging. Finally, we measured forest re-growth after the BBSNP
was declared a national park, to test whether the current BBSNP
promoted forest re-growth.
2.
Study area
Our study area extends for 220 km along the South Barisan
Mountains in southwest Sumatra (Fig. 1), and encompasses
1.17 million ha of land stretching across Sumatra’s southern
Fig. 1 – Patterns of deforestation in a 1.17 million ha area in southwest Sumatra, showing the reduction and fragmentation
of the BBSFL from 1972 to 2002, the logging trail network (mostly active during the period 1972 to 1985), and forest
re-growth in the BBSNP for the period 1985 to 2002. The data are overlaid on a DEM and the 2002 LANDSAT ETM+
imagery with bands 3 (red), 4 (NIR) and 5 (SWIR) combined to generate a false colour background.
provinces of Lampung, Bengkulu, and South Sumatra. Its
southern and western sides are bordered by a 537 km coastline
with the Indian Ocean, while its northern and eastern sides
are defined by the limits of satellite coverage and limits of negligible cloud cover. The topography ranges in elevation from
sea level to 2200 m asl. Rainfall is generally high, between
3000 and 4000 mm per year. In 1972, the study area comprised
one large contiguous area of forest known as, the Bukit Barisan Selatan Forest Landscape (BBSFL). The BBSFL represents
the maximum available forest area in our study site.
Southern Sumatra’s human population has increased dramatically since the 1930s, and this region began to lose its forests sooner than any other Sumatran region (Benoit et al.,
1989). Nevertheless, the BBSFL still contains one of the largest
remaining tracts of lowland and hill rainforests on Sumatra.
BBSFL is home to at least 118 species of mammal, 425 species
of bird, 45 amphibian and reptile species, and 649 species of
higher plant. These include large threatened mammals, such
as Sumatran tigers (Panthera tigris), Asian elephants (Elephas
maximus) and Sumatran rhinoceroses (Dicerorhinus sumatrensis) (O’Brien and Kinnaird, 1996). To conserve the fauna, flora
and watershed functions of the BBSFL the Dutch colonial
power established three reserves in southwest Sumatra during the 1930s. These comprised: the 324,494 ha South Sumatra
I Nature Reserve (SSINR), the 47,782 ha Gunung Raya Wildlife
Sanctuary (GRWS), and a 256,620 ha network of Hydrological
Reserves (HR). In 1984, the South Sumatra I Nature Reserve
was declared as the Bukit Barisan Selatan National Park
(BBSNP). In 2004, the BBSNP was declared a World Heritage site
by UNESCO (decision 28COM 14B. 5).
red, near infrared, and short wave infrared on the ETM+ data
(bands 3, 4 and 5), and red and near infrared on the MSS data
(bands 2, 3 and 4). The red and infrared bands were scattered
less by the atmosphere than the blue and green bands (Born
and Wolf, 1975) and therefore produced the highest available
contrast between forest and non-forest.
We interpreted the imagery as a temporal progression of
forest change to identify areas of deforestation. We defined
‘Deforestation’ as the complete removal of forest cover over
an area equivalent to P1 ha. We defined the ‘forest’ class as
non-modified forest areas of old-growth vegetation dominated
by closed-canopy tree cover (P50%). ‘Non-forest’ comprised
human settlements, Imperata cylindrica grasslands, and regrowth areas dominated by shrubs and young forest trees,
paddy fields and tree crops. Tree crops included coffee and
pepper gardens, cinnamon, coconut and oil palm plantations,
orchards and an old-growth agro-forest dominated by Shorea
javanica a species of Dipterocarp (Michon and de Foresta, 1992).
We used a gaussian Maximum Likelihood Classification
(MLC) algorithm (Schowengerdt, 1997) to generate a classification of the study area into one forest class and one non-forest
class. We filtered classification results to remove clumps of
pixels whose corresponding size on the earth surface was
<1 ha. We edited classification results visually by on-screen
digitizing in areas of haze and in areas of high topographic
complexity, where the MLC algorithm often produces misclassification errors. We digitized logging trails manually by
on-screen digitizing. Small-scale logging by local communities lacking heavy equipment remained undetected.
3.
Methods
3.1.
Mapping forest loss
3.1.1.
Time series satellite data
We employed the NASA SRTM Digital Elevation Model (DEM)
(Rabus et al., 2003) to categorise forest types by elevation
and slope. We placed forest areas into four elevation-based
categories using the altitudinal zoning system proposed by
van Steenis (1961): lowland: 0–500 m; hill: 500–1000 m; lower
montane: 1000–1500 m; and, upper montane: 1500–2200 m
asl. Forest areas were further categorised into four slopebased forest classes, using Indonesia’s Ministry of Forestry
classification system: 0–8.5°, 8.5–24°, 24–45°, +45°.
The BBSNP boundary was obtained from the BBSNP Office
at a scale of 1:25,000, and corrected in the field with GPS, by a
team from the BBSNP Office, the Wildlife Conservation Society (WCS) and WWF. The GRWS and HR boundaries were obtained from the Ministry of Forestry at a scale of 1:250,000, but
were rescaled to 1:25,000 in selected areas by the International Centre for Research in Agro forestry (ICRAF) and
WCS, again using GPS.
3.1.3.
To generate maps of changing forest cover across the study
area, we acquired LANDSAT scenes for the years 1972 (MSS),
1985 (MSS) and 2002 (ETM+), all of which had negligible (less
than 3.5%) cloud cover. The 1985 imagery was used to calculate
deforestation rates during the BBSNP’s pre- and post-establishment periods. We geo-referenced the 2002 scenes to the
ground with Ground Control Points (GCPs) collected in the field
by the World Wildlife Fund (WWF) using a handheld Global
Positioning System (GPS) set in the Universal Transverse Mercator (UTM) projection, Zone 48 South. Then, we matched
the 1972 and 1985 scenes in x- and y- to the 2002 scenes
through a second order polynomial co-registration technique
(Schowengerdt, 1997). The spatial precision obtained was
smaller than one pixel. We obtained additional cloud-free
LANDSAT and aerial imagery for 1976, 1978, 1989, 1994, 1997
and 2000 in areas where large-scale mechanised logging had
occurred to map the spatial distribution of logging trails (truck
roads). The locations of these areas were identified from maps
and reports of old timber concessions bordering BBSNP (Meijer,
1975) and by interviewing long-serving BBSNP staff.
3.1.2.
Classification strategy
We identified forest and non-forest by analysing the reflection
of sunlight from the Earth’s surface in three spectral regions:
3.1.4.
Topography and PA boundary
Accuracy assessment
We employed aerial photography and high-spatial resolution
IKONOS imagery as reference information to validate the
accuracies of the LANDSAT MSS- & TM-based 1972 and 2002
forest maps. The aerial photograph was acquired in 1976 by
the Indonesian air force. The IKONOS imagery was acquired
in 2002 by the IKONOS-2 satellite. The 1976 and 2002 reference
imagery covered a 43,000 ha and 27,000 ha area, respectively
encompassing the primary forests over the full topographic
range of BBSNP and adjacent agricultural areas. The forest
area on the reference imagery was digitized manually using
on-screen digitizing technique to create a reference forest
map for year 1976 (in 1:25,000 scale) and for year 2002 (in
1:3,600 scale) directly comparable to the LANDSAT MSS- &
TM-based 1972 and 2002 forest maps (in 1:108,000 scale).
We generated two confusion matrices (Aronoff, 1982; Foody, 1992) by assigning either forest or non-forest classes to (i)
220 points on the LANDSAT-MSS 1972 forest map and on its
reference map (aerial photography-based) and to (ii) 145
points on the LANDSAT-TM 2002 forest map and its reference
map (IKONOS -based). These points were chosen randomly,
but with a separation distance of 1000 m between two points
to minimize the effects of spatial autocorrelation (Koening,
1999). From each confusion matrix, we calculated the overall
accuracy, po and the kappa coefficient, k (Cohen, 1960). We
could not validate the accuracy of the 1985 forest map because we did not possess reference information for that year.
3.1.5.
Forest fragmentation
To provide ecologically meaningful measures, we extracted
fragmentation statistics only on forest fragments >an arbitrarily chosen threshold of 100 ha (McGarigal and Marks, 1994).
We measured the degree of isolation of each fragment by
calculating the Nearest Neighbour Distance (NND) as an
edge-to-edge distance from one fragment to its nearest neighbour. We measured the degree of roundedness of each fragment by calculating the ratio, R between the minor and
major axes of a fitted ellipse having equal area. We measured
the proportion of forest area for each fragment in low and
high altitudinal forest by calculating the ratio between the
areas of forest at low altitude (<1000 m asl) and at high altitude (>1000 m asl). Likewise, we computed the ratio between
the area of forest on gentle slopes (<16.5°) and on steep slopes
(>16.5°).
As forest fragments sometimes fall across areas with different protection status, we ruled that a given fragment belonged to that particular protection status only when the
latter enclosed more than 50% of the fragment area. Conversely, when more than 50% of a fragment fell into unprotected land, then that fragment was considered as
unprotected. This rule enabled us to quantify the number of
forest fragments by status of PA, in order to calculate their
associated summary fragmentation statistics.
3.2.
Mapping re-growth
While we could map forest loss and logging trails consistently
across the study area with satellite imagery, we could not
map forest re-growth by satellite alone because this land cover type produces similar reflectance values to those generated
by smallholder tree crops, and vice versa (King, 2002). Therefore we combined our forest maps with field surveys to measure the amount of re-growth within the current BBSNP after
it was declared a national park.
3.2.1.
Satellite-based forest maps
The LANDSAT-based forest maps derived for 1972, 1985 and
2002 helped categorise the encroachments into BBSNP (defined as those areas within BBSNP where forest has been removed since 1972). These encroachments were classified as
either ‘active’ or ‘inactive’, based on whether or not they
had expanded in size during BBSNP’s post-establishment period, from 1985 to 2002. We predicted a predominance of regrowth areas and absence of agricultural land in the ‘inactive’
encroachments, but a predominance of agricultural land, and
absence of re-growth in the ‘active’ encroachments. We subsequently verified our satellite-based assumptions with field
surveys.
3.2.2.
Field evaluation of forest re-growth in BBSNP
From September 2004 to May 2005, three teams from the
BBSNP Office and WCS recorded which land use types have
substituted natural forests within ‘active’ and ‘inactive’
encroachments into BBSNP. The forest edge and BBSNP
boundary marked the limit of any areas of encroachment.
Each team sampled agricultural and non-agricultural areas
of land by walking along a line transect placed in each area
of encroachment. Within non-agricultural areas, the teams
sampled re-growth areas, grassland areas as well as recent
clearings. The teams recorded this vegetation change in
nearly all the encroachments to capture the entire spatial variability of forest conversion processes inside the park. Because
of time constraints, the surveyors were prevented from measuring vegetation parameters using tools, as used in smaller
scale, rigorous forest inventories (Husch et al., 1993). We
placed the pre-determined line transect across the full topographic gradient of the encroachment to avoid the selective
sampling of vegetation in more accessible areas, and we
avoided positioning it along footpaths, to encourage surveyors
to sample all the vegetation types within the encroachment.
Each field team carried a LANDSAT-based 108,000 scale
map of the surveyed encroachment area under investigation.
The map displayed the forest edge and BBSNP boundary to
mark the border of the surveyed encroachment, topographic
contour lines, road, village and river networks overlays and
the position of the pre-determined line transect. The teams
used a compass and handheld GPS unit to follow the direction
of the line transect and a digital camera to photograph the
vegetation. While walking each transect, the surveyors captured a GPS reading at the transition from agriculture to
non-agriculture and from areas of re-growth to grasslands
or to recent clearing (and vice-versa), and took a picture of
the landscape. This surveying method provided a length measurement and photograph of every homogeneous stand of
vegetation and land use type encountered along the line transect. The GPS point data and associated attributes were subsequently entered in a Geographic Information System (GIS)
and converted to a length measurement, to partition the line
transect into segments of varying lengths. Each segment had
vegetation and land use attribute and an associated photograph. The photograph was subsequently referred to, to ensure the three teams provided consistent results.
4.
Results
4.1.
Accuracy results
We generated a classification of the study area into forest and
non-forest categories for 1972, 1985 and 2002. The overall
accuracy for the 1972 and 2002 forest/non-forest classification
is po1972 = 95.0% and po2002 = 95.8%, with kappa statistics of
k1972 = 0.86 and k2002 = 0.88, respectively. We believe the classification accuracy for 1985 is as good as for 1972, given that the
same sensors and processing methods were employed across
each classification.
4.2.
Deforestation patterns across the study area
Of the 1.17 million ha study area, over half (692,850 ha) was
covered in natural forest in 1972 (Table 1, Fig. 1). This forest
area constituted one single, large contiguous area of forest,
the BBSFL. Lowland forests, 6500 m asl, and hill forests,
>500 m and 61000 m asl, accounted for 40% and 39% of this
forest ecosystem, respectively.
By 2002, the overall size of BBSFL had reduced by nearly
half (Table 1, Fig. 1), at an average rate per original forest cover
of 1.69% y 1. The rates of loss of lowland, hill and lower montane forest areas were similar, averaging 1.7% y 1, 1.8% y 1
and 1.4% y 1, respectively (Fig. 2). In contrast, the rate of loss
of upper montane forests was much lower at 0.2% y 1. Deforestation rates increased from 0.2% y 1 on extreme slopes of
+45°, to 1.9% y 1 on gentle slopes <8.5°. Overall, 84% of the forest area lost across the BBSFL was of lowland and hill forest
<1000 m asl. However, this loss did not alter the proportion
of different forest types by elevation (Fig. 3).
The loss of forest cover resulted in the breaking up of
BBSFL into smaller fragments. By 2002, BBSFL comprised
1094 forest fragments, but the size of these fragments was
heavily skewed. Approximately 45% of the fragments were
<2.7 ha, while the three largest fragments were 17,673 ha,
120,000 ha and 159,000 ha, respectively. Of the 1094 fragments, 32 fragments were >100 ha. The NND between these
32 fragments ranged from 0.06 km to 9.88 km (Table 2), and
some 72% of the fragments were separated by more than
1 km. The roundedness (R) values of the forest fragments ranged from 0.21 to 0.91. Approximately 19% and 9% of the fragments were either round (R P 0.8), or highly elongated
(R 6 0.3), respectively, and the largest forest fragment had
the most elongated shape (R = 0.21). Approximately 69% and
56% of the fragments contained a higher proportion of forest
area at higher altitudes and on steep slopes, respectively (Table 2).
4.3.
Deforestation patterns within PAs
In 1972, the combined area of PAs made up by the former
SSINR, and by GRWS and HR protected 75% of the BBSFL (Table 1, Fig. 1). In the subsequent 30-year period, unprotected
forests have reduced by 145,264 ha, representing an 84% loss
of original forest cover, at an average rate of 2.86% y 1. In
GRWS, forests have shrunk by 28,696 ha, representing an
81% loss of forest cover, at an average rate of 2.74% y 1, while
forests in HR have reduced by 113,105 ha, representing a 62%
loss of forest cover, at an average rate of 2.13% y 1. In contrast, the forest area of the current BBSNP has reduced by
57,344 ha, representing a 19% loss of forest cover, at an average rate of 0.64% y 1. As with the deforestation pattern across
Table 1 – Statistics for losses in forest cover of the BBSFL from 1972 to 1985 and from 1985 to 2002, shown according to
protected status
Forest categories
BBSFL
Area of PA (ha)
Forest cover in 1972 (ha)
Lowland forest
Hill forest
Lower montane forest
Upper montane forest
All elevations
SSINR/BBSNP
HR
GRWS
Unprotected
—
324,494
256,620
47,782
—
279,983
268,545
135,061
9,261
692,850
124,297
120,626
56,188
2901
304,012
39,409
86,551
50,573
4,124
180,657
648
18,544
14,393
1,977
35,562
115,629
42,824
13,907
259
172,619
Forest cover in 1985 (ha)
Lowland forest
Hill forest
Lower montane forest
Upper montane forest
All elevations
201,946
168,575
97,904
9,193
477,618
112,701
110,682
52,397
2901
278,681
15,785
32,209
31,187
4,095
83,276
205
7,281
9643
1,938
19,067
73,255
18,403
4,677
259
96,594
Forest cover in 2002 (ha)
Lowland forest
Hill forest
Lower montane forest
Upper montane forest
All elevations
133,475
126,095
80,043
8,828
348,441
105,656
92,310
45,851
2851
246,668
12,133
24,616
26,733
4,070
67,552
0
288
4,793
1,785
6866
15,686
8,881
2,666
122
27,355
Forest area reduction 1972–2002 (ha)
Lowland forest
Hill forest
Lower montane forest
Upper montane forest
All elevations
146,508
142,450
55,018
433
344,409
18,641
28,316
10,337
50
57,344
27,276
61,935
23,840
54
113,105
648
18,256
9,600
192
28,696
99,943
33,943
11,241
137
145,264
Rate of forest loss 1972–2002 (% year
1
)
1.69
0.64
2.13
2.74
2.86
Annual rates of forest loss(%/y ) per elevation
5.0
Annual rates of forest loss (%y ) per slope angle
5.0
-1
4.0
0-500 m
500-1000 m
1000- 1500 m
>1500 m
3.0
2.0
1.0
0.0
BBSFL
4.0
BBSNP
HR
GRWS
Unprotected
0-8.5˚
8.5-24 ˚
24-45 ˚
-1
>45 ˚
3.0
2.0
1.0
0.0
BBSFL
BBSNP
HRG
RWS
Unprotected
Fig. 2 – Deforestation rates across BBSFL from 1972 to 2002, arranged by PA and through a gradient of elevation and slope
angle.
the wider BBSFL, the rates of forest loss in the PAs were highest in the lowland, hill and lower montane forests, and rates
of loss increased with decreasing slope angle (Fig. 2). This loss
of forest cover significantly altered the proportional area of
different forest types by elevation in the GRWS and HR, where
the dominant forest type shifted from hill forest to lower
montane forest (Fig. 3).
Fragmentation statistics also differed across PAs of different status (Table 2). The GRWS contained the fewest forest
fragments (n = 3) while the HR had the most fragments
(n = 16). The mean fragment size decreased with increasing
rates of forest loss (Tables 1 and 2). Fragments in the HR
and GRWS had the highest mean NND, of 3.00 km and
2.82 km, respectively. In contrast, the current BBSNP had the
shortest mean NND of 1.25 km. The HR and GRWS also contained a higher proportion of fragments at high altitudes
and on steep slopes than did the BBSNP.
4.4.
Deforestation and logging rates during BBSNP’s preand post-establishment periods
During its pre- (1972–1985) and post-establishment (1985–
2002) periods, the forests within the former SSINR and current
BBSNP, have reduced at similar rates relative to the original
(1972) forest area, and rates of loss have averaged 0.65% y 1
and 0.63% y 1, respectively (Fig. 4).
The logging trail network extended 123 km in the southern and western sections of the former SSINR, during the
pre-establishment period of BBSNP (1972–1985). This network extended 29 km in the western and northern sections
of BBSNP during its post-establishment period (1985–2002)
(Fig. 4).
4.5.
Re-growth during BBSNP’s post-establishment period
Over the thirty years, an area of 57,344 ha of primary forest
has been removed within the boundary of the BBSNP, with
48,734 ha and 8610 ha categorised as ‘active’, and ‘inactive’,
respectively. Most ‘inactive’ encroachments were the subject
of forced evictions in 1986, following the government’s pledge
to intensify the protection of its PAs at the 1982 Bali World
Parks congress.
The total transect length extended 365.3 km, with
342.7 km and 22.6 km in ‘active’ and ‘inactive’ encroachments, respectively.
Agricultural land occurred along 57.9% and 23.1% of the total lengths of ‘active’ and ‘inactive’ encroachments, respectively (Fig. 5).
In ‘inactive’ encroachments, re-growth occurred along
76.9% of the total transect length. This re-growth was dominated by tall shrubs and young forest trees characteristic of
secondary forest and sometimes bamboo or wild ginger (Zingiberaceae). In ‘active’ encroachments, re-growth occurred
along 34.5% of the total transect length. This re-growth was
dominated by unattended coffee plantations and small
shrubs characteristic of fallow vegetation.
100
90
BBSFL
Chi-Square = 2.468
0.5 < p < 0.75
80
70
1972
2002
60
50
40
30
20
10
80
60
50
40
30
20
10
0
Lowland
Hill
Lower
montane
Upper
montane
Lowland
70
Hill
Lower
montane
Upper
montane
100
HR
Percentage (%) of total forested area
Percentage (%) of total forested area
100
80
Chi-Square = 0.505
0.95 < p < 0.975
1972
2002
70
0
90
BBSNP
90
Percentage (%) of totalforested area
Percentage (%) of total forested area
100
Chi-Square = 13.65
0.001 < p < 0.005
1972
2002
60
50
40
30
20
10
0
GRWS
90
80
Chi-Square = 141.88
0.001 < p
1972
2002
70
60
50
40
30
20
10
0
Lowland
Hill
Lower
montane
Upper
Montane
Lowland
Hill
Lower
montane
Upper
Montane
Fig. 3 – Proportions of different forest types by elevation, shown for each of the BBSFL, BBSNP, GRWS and HR, in 1972 and
2002.
Table 2 – Summary of fragmentation statistics within areas of different protection status across the BBSFL in 2002
Fragmentation statistics in 2002
Area of PA (ha)
No. of fragments (>80 ha)
Fragment size (ha)
Mean
Min/Max
Nearest neighbour distance, NND (km)
Mean
Min/Max
No. of fragments
<1000 m ; >1000 m
<16.5°; >16.5°
BBSFL
SSINR/ BBSNP
HR
GRWS
Unprotected
—
32
324,494
7
256,620
16
47,782
3
—
6
11,012
120/158,568
41,300
169/158,568
3480
121/16,361
1587
1325/2084
472
120/973
2.41
0.06/9.88
1.25
0.06/4.43
3.00
0.15/9.88
2.82
1.79/4.14
1.99
0.15/6.74
10/22
14/18
4/3
6/1
2/14
4/12
0/3
0/3
4/2
4/2
These results confirm the predominance of forest regrowth (76.9%) in ‘inactive’ encroachments, and the predominance of agricultural land (57.9%) in ‘active’ encroachments.
However this survey also reports a significant presence of
agricultural land (23.1%) in ‘inactive’ encroachments, and of
fallow vegetation (34.5%) in ‘active’ encroachments.
5.
Discussion
5.1.
Accurate maps of forest cover and change
This study of deforestation over an extensive area of southwest Sumatra has generated maps of forest cover and forest
Annual rate of forest loss ( % y-1)
Logging trails cumulative length (km)
1.00
0.65
0.63
0.50
0.00
1972 -- 1985
1985 -- 2002
Pre-establishment period
Post-establishment period
123
100
50
29
0
1972 - - 1985
1985 - - 2002
Pre-establishment period
Post-establishment period
Fig. 4 – Average annual rates of forest loss and cumulative length of logging trails before and after BBSNP was established.
’Inactive’ encroachments
’Active’ encroachments
(no area expansion since 1985)
(area expansion since 1985)
3%
23%
34%
58%
77%
5%
Recent clearings
Grassland
re-growth
Agriculture
Fig. 5 – The proportion of re-growth and agriculture within ‘active’ and ‘inactive’ encroachments into the BBSNP in 2004–2005.
cover change for 1972, 1985 and 2002 at the very fine spatial
scale of 1 ha with high classification accuracies for both
non-forest and forest classes, appropriate for analysing fine
scale changes in forest cover (Thomlinson et al., 1999). Nevertheless, our accuracy assessments were conducted on just a
6–7% spatial subset, and therefore the accuracies presented
in this study may vary across the study area.
Our deforestation results merit comparison with those
from the earlier study of deforestation over 70% of the area
of BBSNP (Kinnaird et al., 2003), which suggested that forest
cover had reduced by 66,190 ha from 1985 to 1999, representing a loss of forest cover of 2% y 1 during that period. Our
deforestation estimates were calculated within more accurate
park boundaries, endorsed by the Park Office as the official
boundary, and a more accurate classification method, all of
which might explain discrepancies between the rates reported by Kinnaird et al. (2003) and this study.
5.2.
Implications for wildlife conservation
In 1972, the former SSINR, and GRWS and HR, were embedded
in a single, large contiguous area of forest known as the
BBSFL. Wide ranging species of large mammal could easily
disperse from one PA to another through the relative safety
of the dense forest interior. However, the BBSFL had shrunk
by nearly half in 2002, and had broken up into many small for-
est fragments. Furthermore, the existing PAs became spatially
isolated from one another, while their integrity has also been
degraded and fragmented, as has occurred to tropical PAs
world-wide (Sanchez-Azofeifa et al., 2003; Linkie et al., 2004;
Defries et al., 2005; Naughton-Treves et al., 2005; Nepstad
et al., 2006). Increasing isolation and fragmentation may detrimentally affect the movement patterns of wide ranging and
threatened mammals who still live in the BBSFL, such as
Sumatran tigers (P. trigris), Asian elephants (E. maximus), and
Sumatran rhinoceros (D. sumatrensis). In addition, most
remaining forest fragments lie at high altitudes (>1000 m)
and on steep land (>16.5°), rather than in the lowland forest
preferred by most of these species. Widodo and Santiapillai
(1992) found evidence of tiger and elephant presence in the
lowland and hill forests of the GRWS in the 1980 s, but this
small PA, with its highly fragmented forest habitats can no
longer support viable populations of elephants and tigers
(Hedges et al., 2005). An estimated 17 elephants still persist
in the HR, but this small population may not persist in the
long-term (Hedges et al., 2005).
The BBSNP is perhaps the only PA in southwest Sumatra
that still offers good forest habitat for wildlife. It includes
the largest remaining forest patches, which are narrowly disconnected from one another, as well as the largest remaining
expanses of lowland forest on flat areas of land (Tables 1 and
2). It has also partially reclaimed degraded areas, particularly
in the southern peninsular, where large mammal populations
have been recorded. The tiger and elephant populations of
BBSNP were estimated at 40–43 and 498 individuals, respectively in 2001 (O’Brien et al., 2003; Hedges et al., 2005).
Nevertheless, the integrity of the current BBSNP has been
compromised over the past 30 years (see also Kinnaird
et al., 2003). Its buffer forests have almost completely disappeared and its remaining forest fragments are thin and elongated. This renders the wildlife of BBSNP particularly
vulnerable to human pressure, and increases the edge of isolated and remote forest patches compared to a circular park
of the same area. Kinnaird et al. (2003) showed that tigers, rhinoceroses and elephants in BBSNP avoid the forest edge,
probably because of poaching pressure and fear of humans.
Conversely, Hedges et al. (2005) have shown that incidents
of human–elephant conflict at the forest edge have escalated
in recent times, probably because natural habitats have been
fragmented and substituted by farmland, which in turn has
increased the likelihood of people encountering elephants (Sitati et al., 2003).
5.4.
Park effectiveness
Notwithstanding its superior conservation capabilities compared to the other reserves, the BBSNP did not prevent forest
loss after the SSINR was declared a national park in 1984. This
finding was unexpected because the government of Indonesia
implemented radical protection measures after the SSINR
was declared a national park following its pledge to expand
and protect Indonesia’s network of PAs at the 1982 Bali World
Parks congress (Kusworo, 2000). These measures, of cancelling logging permits and evicting rural communities, successfully reduced large-scale mechanized logging by a factor of 4.2
and stabilized some 8610 ha of encroachment, enabling forest
re-growth. Nevertheless, they failed to slow down agricultural
encroachments. The rapid growth of rural population through
spontaneous migrations and the new found rights of the rural
communities to shrug off the impositions of government
agencies since the 1997–1998 economic crises might explain
why the BBSNP failed to inhibit the expansion of agricultural
encroachments. Whether deforestation has accelerated inside the park since 1997–1998, as Curran et al. (2004) have
shown for Gunung Palung National Park in Kalimantan, is
the subject of further research.
6.
Conclusion
This study has showed that the performance of the BBSNP in
Southern Sumatra at conserving forest habitats has achieved
only mixed results. One the one hand, the BBSNP performed
better than its neighbouring landscape, it halted the development of large-scale mechanised logging and to some extent
promoted forest re-growth. On the other hand, it failed to
slow down agricultural encroachments. This finding adds to
the body of evidence that Southeast Asian PAs prevent forest
loss less effectively than Latin American PAs (Curran et al.,
2004; Linkie et al., 2004; Defries et al., 2005).
Broadly, this study demonstrates that the ratio of forest
loss outside versus inside a PA boundary cannot measure
the effectiveness of PAs at preventing forest loss reliably. Such
a measure would have overestimated the effectiveness of the
BBSNP at mitigating forest loss because the park performed
better than its neighbouring landscape, and yet it failed to
slow down agricultural encroachments. Therefore, we question whether the world’s tropical terrestrial PAs truly prevent
forest loss as effectively as claimed, based on three regional
studies that used this measure as their main target (Defries
et al., 2005; Naughton-Treves et al., 2005; Nepstad et al., 2006).
5.3.
Management capabilities among PAs of different
status
Acknowledgements
As expected, the unprotected forests of southwest Sumatra
have been cleared more rapidly than the protected forests.
However, the protected forests of the GRWS and HR have
been cleared almost as rapidly as the unprotected forests
and have become severely fragmented and isolated (Tables
1 and 2), indicating that they have received little protection.
Possibly, the GRWS and HR have been less effective than
the BBSNP at conserving forests because these reserves do
not possess their own management unit while the BBSNP
does.
Our research was a collaborative effort by the Wildlife Conservation Society’s Indonesia Program and the Bukit Barisan
Selatan National Park Office of the Indonesian Ministry of Forestry’s Directorate General of Forest Protection and Nature
Conservation (PHKA). The work was funded by the Wildlife
Conservation Society and the Illegal Logging Response Centre
of the European Union. We thank the head of Bukit Barisan
Selatan National Park, Tamen Sitorus, Wildlife Conservation
Society staff Antonia Gorog, Donny Gunaryadi and Aslan Baco
for their support and guidance to the project. We also thank
Nigel Leader-Williams, Matthew Linkie, Etienne Delattre, Eric
Sanderson, Gari Paoli, Lisa Curran and three anonymous referees, for improving this manuscript. Finally, this work would
not have been possible without the dedication of the field
team: Bambang Pandu Bharoto, Deky Kristiantono, Susilo, Dadan Ramradi, Supriatna, Muhammad Zubair, Asiantori, and
Wakijan.
R E F E R E N C E S
Achard, F., Eva, H.D., Stibig, H.J., Mayaux, P., Gallego, J.,
Richards, T., Malingreau, J.P., 2002. Determination of
deforestation rates of the World’s humid tropical forests.
Science 297, 999–1002.
Aronoff, S., 1982. The map accuracy report: a user’s view.
Photogrammetric Engineering and Remote Sensing 48, 1309–
1312.
Benoit, D., Levang, P., Pain, M., Sevin, O., 1989. Transmigration et
migrations spontanées en Indonesie: Propinsi Lampung.
Departemen Transmigrasi, Jakarta, Indonesia and ORSTOM,
Bondy, France.
Born, M., Wolf, E., 1975. Principles of Optics, fifth ed. Pergamon,
New York.
Bruner, A.G., Gullison, R.E., Rice, R.E., da Fonseca, G.A.B., 2001.
Effectiveness of parks in protected tropical biodiversity.
Science 291, 125–128.
Cohen, J., 1960. A coefficient of agreement for nominal scale.
Educational and Psychological Measurement 20, 37–46.
Curran, L., Trigg, S.N., McDonald, A.K., Astiani, D., Hardiono, M.,
Siregar, P., Caniago, I., Kasischke, E., 2004. Lowland forest loss
in protected areas of Indonesian Borneo. Science 303, 1000–
1003.
Defries, R., Hansen, A., Newton, A.C., Hansen, M.C., 2005.
Increasing isolation of protected areas in tropical forests over
the past twenty years. Ecological Applications 15 (1), 19–26.
Foody, G.M., 1992. On the compensation for chance agreement in
image classification accuracy assessment. Photogrammetric
Engineering and Remote Sensing 58, 1459–1460.
FWI/GFW, 2002. The state of the forest: Indonesia. Forest Watch
Indonesia, and Washington DC: Global Forest Watch, Bogor,
Indonesia.
Hedges, S., Tyson, M.J., Sitompul, A.F., Kinnaird, M.F., Gunaryadi,
D., Aslan, 2005. Distribution, status, and conservation needs of
Asian elephants (Elephas maximus) in Lampung Province,
Sumatra, Indonesia. Biological Conservation 124, 5–48.
Husch, B., Miller, C.I., Beers, T.W., 1993. Forest Mensuration, Third
ed. Krieger Publishing Company, Malabar, FL.
Jepson, P., Jarvie, J.K., MacKinnon, K., Monk, K.A., 2001.
Decentralization and illegal logging spell the end for
Indonesia’s lowland forests. Science 292, 859–861.
King, R.B., 2002. Land cover mapping principles: a return to
fundamentals. International Journal of Remote Sensing 23
(18), 3525–3545.
Kinnaird, M.F., Sanderson, E.W., O’Brien, T.G., Wibisono, H.T.,
Woolmer, G., 2003. Deforestation trends in a tropical
landscape and implications for endangered large mammals.
Conservation Biology 17, 245–257.
Koening, W.D., 1999. Spatial autocorrelation of ecological
phenomena. Trends in Ecology and Evolution 14, 22–26.
Kusworo, A., 2000. Perambah hutan atau kambing hitam? Potret
sengketa kawasan hutan di Lampung. Pustaka Latin, Bogor.
Linkie, M., Smith, R.J., Leader-Williams, N., 2004. Mapping and
predicting deforestation patterns in the lowlands of Sumatra.
Biodiversity and Conservation 13, 1809–1818.
Liu, J., Linderman, M., Ouyang, Z., An, L., Yang, J., Zhang, H., 2002.
Ecological degradation in protected areas: the case of Woolong
Nature Reserve for giant pandas. Science 292, 98–101.
McGarigal, K., Marks, B.J., 1994. FRAGSTATS, spatial pattern
analysis program for quantifying landscape structure. Forest
Science Department, Oregon State University.
Meijer, W., 1975. Report on a fact-finding tour aimed towards a coordinated development of forest resources in Indonesia.
Thomas Hunt Morgan School of Biological Sciences, University
of Kentucky.
Michon, G., de Foresta, H., 1992. Complex agroforestry systems
and conservation of biological diversity. Agroforestry in
Indonesia: a link between two worlds. In: Kheong Y.S., Win L.S.
(Eds), Harmony with Nature: An International Conference on
the Conservation of Tropical Biodiversity, The Malayan Nature
Journal (golden Jubilee issue), Kuala Lumpur, Malaysia.
Naughton-Treves, L., Buck Holland, M., Brandon, K., 2005. The role
of protected areas in conserving biodiversity and sustaining
local livelihoods. Annual Review of Environment and
Resources 30, 219–252.
Nepstad, D., Schwartzman, S., Bamberger, B., Santilli, M., Ray, D.,
Schlesinger, P., Lefebvre, P., Alencar, A., Prinz, E., Fiske, G., Rolla,
A., 2006. Inhibition of amazon deforestation and fire by parks
and indegenous lands. Conservation Biology 20 (1), 65–73.
O’Brien, T.G., Kinnaird, M.F., 1996. Birds and mammals of the
Bukit Barisan Selatan National Park, Sumatra. Indonesia. Oryx
30, 207–217.
O’Brien, T.G., Kinnaird, M.F., Wibisono, H.T., 2003. Crouching
tigers, hidden prey: Sumatran tiger ad prey population in a
tropical landscape. Animal Conservation 6, 131–139.
Rabus, B., Eineder, M., Roth, A., Balmer, R., 2003. The shuttle radar
topography mission-a new class of digital elevation models
acquired by spaceborne radar. Photogrammetry and Remote
Sensing 57, 241–262.
Sanchez-Azofeifa, A., Daily, G.C., Pfaff, A.S.P., Busch, C., 2003.
Integrity and isolation of Costa Rica’s national parks and
biological reserves: examining the dynamics of land-cover
change. Biological Conservation 109, 123–135.
Schowengerdt, R.A., 1997. Models and Methods for Image
Processing. Academic Press, London.
Sitati, N.W., Walpole, M.J., Smith, R.J., Leader-Williams, N., 2003.
Predicting spatial aspects of human-elephant conflict. Journal
of Applied Ecology 40, 667–677.
Steenis, C.G.G.J.van., 1961. An attempt towards an explanation of
the effect of mountain mass elevation. Proceedings of the
Royal Netherlands Academy of Arts and Sciences C64, 435–442.
Thomlinson, J.R., Bolstad, P.V., Cohen, W.B., 1999. Coordinating
methodologies for scaling land-cover classifications from sitespecific to global: steps towards validating global map
products. Remote Sensing of the Environment 70, 16–28.
Widodo, S.R., and Santiapillai, C., 1992. Conservation of Sumatran
Tiger (Panthera tigris sumatrae) in Indonesia. Sumatran tiger
population and habitat viability analysis workshop, Padang,
West Sumatra, Indonesia.